On Fri, Nov 23, 2018 at 07:12:32AM +0100, Landry Breuil wrote:
> On Thu, Nov 22, 2018 at 06:10:44PM +0100, Landry Breuil wrote:
> > On Thu, Nov 22, 2018 at 11:57:53AM -0500, George Koehler wrote:
> > > On Wed, 21 Nov 2018 08:40:06 +0100
> > > Landry Breuil <[email protected]> wrote:
> > > 
> > > > check the PFRAG.${ARCH} mechanism, check x11/vlc or lang/rust
> > > > for examples.
> > > 
> > > That was the right hint!  In the attached diff for databases/mariadb,
> > > I added PFRAG.mroonga-{server,tests} and caused it to install mroonga
> > > on little endian but not big endian:
> > > 
> > > .if ${PROPERTIES:Mle}
> > > PKG_ARGS+=-Dmroonga=1
> > > .else
> > > PKG_ARGS+=-Dmroonga=0
> > > .endif
> > > 
> > > This diff caused `make repackage` to succeed on my big-endian powerpc
> > > machine.  Then I did `make update FORCE_UPDATE=hard` (because I didn't
> > > change REVISION).  I also built mariadb with the same diff on
> > > little-endian amd64, where mroonga is in the packages.  There is no
> > > big-endian mroonga because WRKSRC/server/mroonga/CMakeLists.txt skips
> > > mroonga on big-endian platforms.  `make test` in databases/mariadb
> > > has "2 tests failed out of 56" on both powerpc and amd64.
> > > 
> > > The diff is large, because PFRAG.mroonga-tests has many files, so I
> > > used gzip(1).  (I forgot to use gzip on the much larger
> > > mariadb-server-10.0.37v1.log in my original mail.)
> > 
> > The diff reads fine to me. Usually its better to cc maintainer when
> > there's one.. and you'll be missing a REVISION bump i guess.
> 
> Fwiw, now that cmake & pcre have pending diffs in my tree, mariadb
> failed to package on sparc64 like on macppc. So this diff will surely
> help mariadb there, and unblock everything depending on it.
> 
> But to make progress we need feedback from maintainer :)

Another 'simpler/smaller' fix would be to re-disable the mroonga storage
thing via CONFIGURE_ARGS+=-DWITHOUT_MROONGA_STORAGE_ENGINE=1, which was
removed (and thus enabled it) in
https://github.com/openbsd/ports/commit/2b6f8eb5dade2669b0bb46addc431eec2ce171c3.

Brad, your call.

Landry

Reply via email to